An unmanned aerial vehicle spraying method based on path planning

Abstract

The application relates to the field of information technology and discloses a path planning-based unmanned aerial vehicle spraying method, which comprises the following steps: acquiring overhead line image data, segmenting the line image, and extracting key part images of the overhead line; matching the extracted key part images with standard key part images to judge the key part state; establishing a fault diffusion prediction model to predict a fault diffusion area according to a fault position and a fault type; planning an unmanned aerial vehicle circuit spraying path according to the predicted fault diffusion range and the spraying area terrain, and preferentially spraying the fault diffusion area; judging the processing priority of the fault area within the time constraint limit according to the diffusion area range and the limitation of the single flight time of the unmanned aerial vehicle; determining the spraying path, selecting the best nozzle, setting the spraying speed and the spraying pressure according to the fault processing priority. Compared with the prior art, the spraying method reduces the time and cost of manual inspection and fault positioning.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a path planning-based drone painting method. Background Technology

[0002] With the continuous development and increasing intelligence of overhead lines, fault handling and maintenance of overhead lines have become a critical issue. Specifically, fault detection and location of existing overhead lines still rely on manual intervention. This not only wastes a lot of time but also leads to a continuous increase in construction costs. Furthermore, accurately understanding the status of the line is crucial for maintenance work. Traditional monitoring methods may not be able to obtain accurate status information of the line in real time or can not monitor different parts of the line in detail. How to develop high-precision line status monitoring technology to obtain various parameters and status information of the line in real time and accurately is an urgent problem to be solved.

[0003] Furthermore, with the application of technologies such as drones and sensors, the amount of data generated by overhead line maintenance is constantly increasing. It is necessary to efficiently process and manage this large-scale data and extract useful information to guide and optimize maintenance work. Traditional maintenance plans and prioritization arrangements often lack planning and scientific rigor, leading to extended maintenance cycles and wasted resources. It is necessary to develop efficient maintenance plans and prioritization arrangements based on factors such as the importance of the line, the urgency of the fault, and the availability of maintenance resources.

[0004] Furthermore, overhead line maintenance is often carried out in various complex environmental conditions, such as harsh weather and high-altitude areas. How to ensure that drones can operate normally under various environmental conditions and guarantee their reliability and stability is also a problem that needs to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, the purpose of this invention is to provide a path planning-based UAV spraying method to improve the problem that fault detection and location of overhead lines still rely on manual intervention.

[0006] Based on this, the present invention provides a path planning-based UAV painting method, which includes the following steps:

[0007] S1. Obtain image data of overhead lines, segment the line images, and extract images of key parts of the overhead lines.

[0008] S2. Match the extracted key part images with the standard key part images to determine the state of the key parts;

[0009] S3. If there is a fault in a critical part, record the fault type and establish a three-dimensional mapping model, mark the fault location coordinates, and obtain the obstacle avoidance space distance around the fault.

[0010] S4. Establish a fault propagation prediction model to predict the fault propagation area based on the fault location and fault type.

[0011] S5. Based on the predicted fault spread range and the terrain of the spraying area, plan the UAV circuit spraying path and prioritize spraying the fault spread area.

[0012] S6. Based on the range of the diffusion area and the limitation of a single flight time of the UAV, determine the processing priority of the fault area within the time constraint.

[0013] S7. Determine the spraying path based on the fault handling priority, select the best nozzle, and set the spraying speed and spraying pressure.

[0014] S8. When using a drone for live-line spraying, adjust the paint viscosity according to the spraying pressure and speed, check whether the nozzle is unobstructed, and after adjusting the paint, use a matching drone for live-line spraying.

[0015] In some embodiments of this application, step S1 includes:

[0016] S11. Obtain an image dataset containing overhead lines by taking on-site photos with a drone. This image dataset contains images of overhead lines under different preset scenes and different preset lighting conditions.

[0017] S12. Denoise, resize, and enhance the contrast of the acquired image;

[0018] S13. Annotate the images to identify the locations of key components such as towers, insulators, and jumpers, and divide the labeled dataset into a training set and a validation set.

[0019] S14. Train the model using the training set and optimize the model parameters using cross-entropy loss;

[0020] S15. Segment the overhead line images in the test set, label the segmented areas, and extract the segmented areas belonging to towers, insulators, and jumpers based on the label information.

[0021] In some embodiments of this application, step S2 includes:

[0022] S21. Set a threshold to distinguish the state of key parts and obtain standard key part images, including sample images in normal state and sample images in fault state, and use the extracted key part images as the key part images to be detected.

[0023] S22. Through image preprocessing, the standard key part image and the key part image to be detected are normalized and adjusted to the same size and contrast.

[0024] S23. Input the preprocessed standard key part images into the convolutional neural network, output the part state category, and train the network model through backpropagation.

[0025] S24. Input the image of the key part to be detected into the trained convolutional neural network model, set a threshold based on the matching similarity to determine the state of the key part, and output the state classification result of the key part to be detected. If the state of the key part is fault, output the fault prompt such as fault type, key part name and key part coordinates.

[0026] In some embodiments of this application, step S3 includes:

[0027] S31. Using the camera and lidar of the UAV, acquire images of the surrounding area of ​​the route, establish a three-dimensional mapping model of the terrain and building obstacles around the route, and acquire the GPS position data of the UAV and the relative positional relationship between the UAV and the target on the route.

[0028] S32. Based on the location of the UAV, obtain the coordinate information of the faulty part, use the three-dimensional mapping model to calculate the coordinates of the faulty part on the line, and convert the coordinate information of the faulty part into the absolute coordinates of the line.

[0029] S33. Mark the fault coordinates in the three-dimensional mapping model, and obtain the three-dimensional data of the surrounding environment of the line based on the absolute coordinates of the fault on the line.

[0030] In some embodiments of this application, step S4 includes:

[0031] S41. Obtain historical fault data, including the location coordinates, type, environmental conditions, and spread range of the fault;

[0032] S42. Preprocess the data, filter out abnormal data, and perform normalization.

[0033] S43. Extract features that affect fault propagation, such as fault type, current parameters and climate conditions, from the historical fault data, and construct feature vectors for fault samples.

[0034] S44. Establish a fault propagation prediction model, input the obtained fault characteristics, predict the fault propagation area, and output the predicted fault propagation area.

[0035] S45. If the predicted fault propagation area does not match the actual situation, obtain the sample data after the propagation and iteratively optimize the model.

[0036] S46. Mark the predicted fault propagation area on the three-dimensional mapping model.

[0037] In some embodiments of this application, step S5 includes:

[0038] S51. Use a fault propagation prediction model to predict the fault propagation area, and determine the boundary coordinates of the fault propagation area based on the prediction results.

[0039] S52. Use a three-dimensional mapping model to analyze the distribution of obstacles in the spraying area;

[0040] S53. Based on the obstacle distribution analysis of the fault propagation area and the spraying area, plan the UAV circuit spraying path;

[0041] S54. Based on obstacle distribution factors, use a fast-expanding random tree to generate the optimal spraying path and use it as the key point of the spraying path to guide the flight and spraying operation of the UAV.

[0042] S55. During the path planning process, the fault spread area is taken as the priority target area for spraying to ensure timely coverage of the fault area.

[0043] S56. Based on the turning radius of the UAV, optimize and smooth the path to obtain the initial path. Determine the turning radius limit of the UAV based on its size, maneuverability, and safety requirements.

[0044] S57. Use spline path interpolation to process the initial path, adjust the control points or parameters of the path, and during the smoothing process, adjust the curvature and turning angle of the path according to the turning radius limit of the UAV to ensure that the UAV turns smoothly without exceeding the limit.

[0045] S58. Establish a convolutional neural network model to analyze the spraying effect. Use known good spraying samples and uncovered area samples as inputs to the convolutional neural network model for supervised learning. The convolutional neural network model outputs the uncovered area.

[0046] S59. After the spraying is completed, input the image after spraying into the trained convolutional neural network model, output the uncovered area, that is, the part that is not fully covered due to insufficient spraying angle, determine whether there is an area that is not covered due to insufficient spraying angle, and if there is an area that is not covered due to insufficient spraying angle, optimize the spraying path.

[0047] In some embodiments of this application, step S6 includes:

[0048] S61. Use the fault propagation prediction model to predict the fault propagation area and determine the boundary coordinates of the fault propagation area.

[0049] S62. Determine the limit on the single flight time of the drone based on its battery capacity, flight speed, and endurance.

[0050] S63. Determine the processing priority of the fault area within the time constraint based on the area of ​​the fault area and the distance from the drone's starting point. The larger the spread area and the closer the distance to the obstacle, the higher the priority.

[0051] S64. Sort the faulty areas according to priority and determine the order of processing;

[0052] S65. Based on the sorted priorities, formulate a processing plan within the time constraint. Areas with higher priorities than the preset priorities are arranged in the first flight painting plan, and areas with lower priorities than the preset priorities are placed in the second flight plan. Based on real-time risk and energy consumption data, dynamically optimize the safety and efficiency of UAV painting operations.

[0053] S66. Based on real-time risk and energy consumption data, the safety and efficiency of UAV spraying operations are dynamically optimized, and a comprehensive risk assessment model is constructed. The risk assessment formula is as follows:

[0054] R = a*(1 / E) + b*W + c*(1 / T) + d*D

[0055] Where R is the risk value, coefficients a, b, c, and d represent the weight of factors, 1 / E represents the reciprocal of the remaining battery energy (the lower the battery level, the higher the risk), W directly represents the wind speed (the higher the wind speed, the higher the risk), 1 / T is the reciprocal of the remaining working time (the shorter the time, the higher the risk), and D represents the distance to the nearest charging station (the farther the distance, the higher the risk).

[0056] S67. Construct an energy consumption prediction model. The energy consumption prediction formula is as follows:

[0057] C = a * P + b * S^2 + y * |V|

[0058] Where C is the energy consumption, a, b, and y are coefficients adjusted according to the equipment characteristics, P represents the spraying power consumption, S^2 represents the square of the flight speed, and |V| is the absolute value of the vertical speed change.

[0059] S68. Combining the outputs of the comprehensive risk assessment model and the energy consumption prediction model, the flight altitude, speed and spraying parameters of the UAV are adjusted in real time. When the risk assessment value exceeds the preset threshold or the energy consumption is higher than the preset value, the flight plan is automatically adjusted.

[0060] S69. By using the drone's sensors, wind speed, battery level, location, and flight time data are acquired in real time. The acquired data is fed back to the comprehensive risk assessment model and energy consumption prediction model in real time. When the risk is higher than the preset risk or the energy is lower than the preset energy, the emergency plan is activated.

[0061] In some embodiments of this application, step S7 includes:

[0062] S71. Based on the determined fault area priority, load the corresponding painting path into the UAV's flight control system;

[0063] S72. Select the appropriate nozzle type for the path line according to the priority area type. If the priority area is a plane, select a plane atomizing nozzle; if it is a curved surface, select a cone nozzle with deflection.

[0064] S73. Determine the nozzle flow rate based on the priority of the fault area;

[0065] S74. Select the optimal nozzle based on nozzle flow rate and nozzle type;

[0066] S75. Calculate the required spraying speed based on the size of the fault area and the nozzle flow rate;

[0067] S76. Determine the required spraying pressure based on the nozzle flow rate and the spraying speed.

[0068] In some embodiments of this application, step S8 includes:

[0069] S81. Based on the spraying pressure and speed parameters and the constraints of nozzle unobstructedness, determine the viscosity range of the coating, perform rotational viscosity testing on the coating, and obtain the viscosity data of the coating.

[0070] S82. Determine whether the paint passes smoothly through the nozzle for spraying. If the nozzle is blocked or not clear, it means that the paint viscosity is too high. Dilute or adjust the paint composition to reduce the paint viscosity. If there is overspraying and paint runoff, it means that the paint viscosity is too low. Thicken or adjust the paint composition to increase the paint viscosity.

[0071] S83. Based on the conductivity requirements during the spraying process, adjust the composition of the coating to make it conductive, and determine whether the adhesion of the coating meets the requirements. If it does not meet the requirements, adjust the composition of the coating to improve the adhesion of the coating.

[0072] S84. Based on the viscosity, conductivity, and adhesion of the coating, determine the final coating formula and select a drone spraying method that matches the viscosity of the coating.

[0073] S85. The formula for adjusting the paint viscosity during UAV spraying based on real-time environmental data and predicting paint behavior is as follows:

[0074] P = a*T + b*H + c*V + d*L;

[0075] Where P represents the predicted spraying parameters, T represents the ambient temperature, H represents the ambient humidity, V represents the wind speed, L represents the viscosity of the coating, and a, b, c, and d are adjustment coefficients.

[0076] S86. Install environmental monitoring equipment on the drone to collect temperature, humidity and wind speed data in real time, and obtain real-time rheological property data of the coating, including viscosity and flowability.

[0077] S87. Using the paint behavior formula, adjust the spraying speed, pressure and nozzle angle based on real-time monitoring data. When the temperature rises or the wind speed increases, adjust the spraying pressure and speed to maintain the uniform distribution of the paint.

[0078] S88. Based on actual spraying test data, calibrate and optimize the parameters in the formula, and update the model regularly to adapt to new coating properties and environments.

[0079] The UAV painting method based on path planning provided in this invention has the following advantages compared with the prior art:

[0080] This invention provides a path planning-based UAV spraying method. Through automated image processing and segmentation technology, it quickly acquires images of key components of overhead power lines, matches them with standard images, and determines the status of these key components, improving the efficiency of maintenance personnel and reducing the time and cost of manual inspections and fault location. Furthermore, this application combines fault location and type information to predict the fault's spread area, helping maintenance personnel to take preventative measures to avoid further escalation. Even further, based on the predicted fault spread range and the terrain of the spraying area, this application utilizes UAVs for circuit spraying, prioritizing the spraying of fault spread areas and optimizing and smoothing the path to improve spraying effectiveness and coverage. In practice, based on the spread area range, UAV single flight time limitations, and real-time risk and energy consumption data, the safety and efficiency of UAV spraying operations are dynamically optimized, and the processing priority of fault areas is determined according to actual conditions, improving operational efficiency. By adjusting the paint viscosity and using a matching UAV for live-line spraying, the safety of the spraying operation is improved. Adjusting the paint viscosity during UAV spraying operations based on real-time environmental data further ensures operational safety and effectiveness. Attached Figure Description

[0081] Figure 1 This is a schematic diagram of a path planning-based drone painting method according to an embodiment of the present invention. Detailed Implementation

[0082] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0083] It should be understood that the terms "before," "after," etc., are used in this invention to describe various types of information, but this information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, "before" information can also be called "after" information, and "after" information can also be called "before" information, without departing from the scope of this invention.

[0084] like Figure 1 As shown, this invention provides a drone painting method based on path planning, which includes the following steps:

[0085] S1. Obtain image data of overhead lines, segment the line images, and extract images of key parts of the overhead lines.

[0086] S2. Match the extracted key part images with the standard key part images to determine the state of the key parts;

[0087] S3. If there is a fault in a critical part, record the fault type and establish a three-dimensional mapping model, mark the fault location coordinates, and obtain the obstacle avoidance space distance around the fault.

[0088] S4. Establish a fault propagation prediction model to predict the fault propagation area based on the fault location and fault type.

[0089] S5. Based on the predicted fault spread range and the terrain of the spraying area, plan the UAV circuit spraying path and prioritize spraying the fault spread area.

[0090] S6. Based on the range of the diffusion area and the limitation of a single flight time of the UAV, determine the processing priority of the fault area within the time constraint.

[0091] S7. Determine the spraying path based on the fault handling priority, select the best nozzle, and set the spraying speed and spraying pressure.

[0092] S8. When using a drone for live-line spraying, adjust the paint viscosity according to the spraying pressure and speed, check whether the nozzle is unobstructed, and after adjusting the paint, use a matching drone for live-line spraying.

[0093] This invention provides a path planning-based UAV spraying method. Through automated image processing and segmentation technology, it quickly acquires images of key components of overhead power lines, matches them with standard images, and determines the status of these key components, improving the efficiency of maintenance personnel and reducing the time and cost of manual inspections and fault location. Furthermore, this application combines fault location and type information to predict the fault's spread area, helping maintenance personnel to take preventative measures to avoid further escalation. Even further, based on the predicted fault spread range and the terrain of the spraying area, this application utilizes UAVs for circuit spraying, prioritizing the spraying of fault spread areas and optimizing and smoothing the path to improve spraying effectiveness and coverage. In practice, based on the spread area range, UAV single flight time limitations, and real-time risk and energy consumption data, the safety and efficiency of UAV spraying operations are dynamically optimized, and the processing priority of fault areas is determined according to actual conditions, improving operational efficiency. By adjusting the paint viscosity and using a matching UAV for live-line spraying, the safety of the spraying operation is improved. Adjusting the paint viscosity during UAV spraying operations based on real-time environmental data further ensures operational safety and effectiveness.

[0094] Optionally, step S1 of this application includes the following steps:

[0095] S11. Obtain an image dataset containing overhead lines by taking on-site photos with a drone. This image dataset contains images of overhead lines under different preset scenes and different preset lighting conditions.

[0096] S12. Denoise, resize, and enhance the contrast of the acquired image;

[0097] S13. Annotate the images to identify the locations of key components such as towers, insulators, and jumpers, and divide the labeled dataset into a training set and a validation set.

[0098] S14. Train the model using the training set and optimize the model parameters using cross-entropy loss;

[0099] S15. Segment the overhead line images in the test set, label the segmented areas, and extract the segmented areas belonging to towers, insulators, and jumpers based on the label information.

[0100] Specifically, in some embodiments of this application, operators can obtain a dataset of 200 images containing overhead power lines through on-site drone photography. The images cover different preset scenes such as cities, rural areas, and mountainous regions, as well as overhead power line images under different preset lighting conditions. The images are then preprocessed, including denoising, resizing, and contrast enhancement. Specifically, a Gaussian filter is used to denoise the images, reducing noise. The images are resized to a uniform size, with width and height adjusted to 512 pixels. Histogram equalization and contrast stretching methods are used to enhance image contrast. Rectangular bounding boxes are used to label the images, marking the locations of towers, insulators, and jumpers. 100 towers, 250 insulators, and 80 jumpers are labeled in the dataset. The dataset is divided into a training set and a validation set, with 80% of the images used for training and 20% for validation. The training set contains 160 images, and the validation set contains 40 images. The U-Net model was trained using 160 images and corresponding labeled data from the training set. Model parameters were optimized using the cross-entropy loss function, and the Adam optimizer was used for training. 100 training epochs were performed, with a batch size of 10 per epoch. The trained U-Net model was then used to segment 40 images from the test set. The segmented image regions were classified according to their labels, and regions belonging to towers, insulators, and jumpers were extracted. In this embodiment, segmented regions belonging to 30 towers, 80 insulators, and 20 jumpers were extracted from the test set.

[0101] Optionally, step S2 of this application includes the following steps:

[0102] S21. Set a threshold to distinguish the state of key parts and obtain standard key part images, including sample images in normal state and sample images in fault state, and use the extracted key part images as the key part images to be detected.

[0103] S22. Through image preprocessing, the standard key part image and the key part image to be detected are normalized and adjusted to the same size and contrast.

[0104] S23. Input the preprocessed standard key part images into the convolutional neural network, output the part state category, and train the network model through backpropagation.

[0105] S24. Input the image of the key part to be detected into the trained convolutional neural network model, set a threshold based on the matching similarity to determine the state of the key part, and output the state classification result of the key part to be detected. If the state of the key part is fault, output the fault prompt such as fault type, key part name and key part coordinates.

[0106] Specifically, in some embodiments of this application, standard images of key components of the line are acquired, including sample images under normal conditions and sample images under fault conditions. Sample images under normal conditions include images of towers, insulators, and jumpers, while sample images under fault conditions include images of fracture, corrosion, or short-circuit faults. The standard key component images and the images of key components to be detected are preprocessed, adjusting the images to the same size, adjusting the width and height to 256 pixels, and enhancing image contrast through histogram equalization. The preprocessed standard key component images are input into a convolutional neural network (CNN) model, which outputs the category of the key component's state, including normal, fracture, corrosion, or short circuit. The CNN model is trained using a backpropagation algorithm, and the model parameters are optimized by minimizing the loss function. The images of key components to be detected are input into the trained CNN model to obtain the classification results of the key component's state. Based on the matching similarity, a threshold is set to determine the state of the key component. If a jumper fracture fault is detected, the fault type is output as fracture, the key component name as jumper, and the coordinate information of the fractured jumper.

[0107] For step S3 of this application, it specifically includes the following steps:

[0108] S31. Using the camera and lidar of the UAV, acquire images of the surrounding area of ​​the route, establish a three-dimensional mapping model of the terrain and building obstacles around the route, and acquire the GPS position data of the UAV and the relative positional relationship between the UAV and the target on the route.

[0109] S32. Based on the location of the UAV, obtain the coordinate information of the faulty part, use the three-dimensional mapping model to calculate the coordinates of the faulty part on the line, and convert the coordinate information of the faulty part into the absolute coordinates of the line.

[0110] S33. Mark the fault coordinates in the three-dimensional mapping model, and obtain the three-dimensional data of the surrounding environment of the line based on the absolute coordinates of the fault on the line.

[0111] Specifically, in some embodiments of this application, the GPS position data of the UAV is (31°23′4″, 45°57′8″), and the relative position of the target on the route is 10 meters directly in front of the UAV. Images of the surrounding area are acquired using the UAV's camera and LiDAR. The images of the surrounding area captured by the UAV camera include a building and a tree. A 3D model of the surrounding terrain and obstacles is established, with the building height set to 20 meters and the tree height set to 10 meters. The coordinates of the fault location are (2, 0, 0), representing the offset of the fault location relative to the UAV's position (2 meters forward, 0 meters to the right, 0 meters upward). The coordinates of the fault location on the route are calculated using the 3D mapping model. The absolute coordinates of the UAV's position are (30, 40, 0), therefore the absolute coordinates of the fault location on the route are (30+2, 40+0, 0) = (32, 40, 0). The fault coordinates are marked in the 3D mapping model, specifically the fault coordinates (32, 40, 0). Based on the absolute coordinates of the fault on the line, 3D data of the surrounding environment is obtained. The 3D data shows that there is a building near the fault location, with the closest point to the fault location being 5 meters away, and trees are 8 meters away from the closest point to the fault location. The obstacle avoidance space distance between the fault location and obstacles is calculated, i.e., the closest distance between the fault location and the surface of the building and tree. The preset obstacle avoidance space distance is 3 meters. According to the calculation, the obstacle avoidance space distance for the building is 5 meters, which is greater than the preset distance, therefore it is not marked as an area of ​​interest. The obstacle avoidance space distance for the tree is 8 meters, which is also greater than the preset distance, so it is not marked as an area of ​​interest either.

[0112] For step S4 of this application, it specifically includes the following steps:

[0113] S41. Obtain historical fault data, including the location coordinates, type, environmental conditions, and spread range of the fault;

[0114] S42. Preprocess the data, filter out abnormal data, and perform normalization.

[0115] S43. Extract features that affect fault propagation, such as fault type, current parameters and climate conditions, from the historical fault data, and construct feature vectors for fault samples.

[0116] S44. Establish a fault propagation prediction model, input the obtained fault characteristics, predict the fault propagation area, and output the predicted fault propagation area.

[0117] S45. If the predicted fault propagation area does not match the actual situation, obtain the sample data after the propagation and iteratively optimize the model.

[0118] S46. Mark the predicted fault propagation area on the three-dimensional mapping model.

[0119] In some embodiments of this application, operators acquire 5000 historical fault data points, including features such as fault location, type, and final spread range. Normalization is performed on numerical features such as temperature, humidity, and corrosion level. Ten features, including fault type, line material, temperature, and humidity, are selected to construct sample vectors. A support vector machine algorithm is used, with an RBF kernel set, and a fault spread prediction model is obtained after training. A newly occurring fault is acquired, located at (120, 30), of type insulator breakage, with a local temperature of 32°C, humidity of 80%, and the faulty line material being aluminum alloy. An input vector is constructed based on the known information, and the fault spread prediction model predicts the fault spread range to be within a 10-meter radius. In actual monitoring, it is found that the fault has spread to a 15-meter radius two days later. To address this prediction deviation, the spread sample data of the fault is collected, added to the training set, and the penalty parameter of the fault spread prediction model is adjusted to 0.8, and the model is retrained. After multiple iterations and optimizations, the prediction accuracy reaches over 90%. Based on the 10-meter prediction range given by the fault propagation prediction model, this range is drawn on the three-dimensional line model to guide subsequent maintenance operations. Anti-corrosion coating is applied in a timely manner within a 15-meter range to prevent the fault from spreading further.

[0120] For step S5 of this application, it specifically includes the following steps:

[0121] S51. Use a fault propagation prediction model to predict the fault propagation area, and determine the boundary coordinates of the fault propagation area based on the prediction results.

[0122] S52. Use a three-dimensional mapping model to analyze the distribution of obstacles in the spraying area;

[0123] S53. Based on the obstacle distribution analysis of the fault propagation area and the spraying area, plan the UAV circuit spraying path;

[0124] S54. Based on obstacle distribution factors, use a fast-expanding random tree to generate the optimal spraying path and use it as the key point of the spraying path to guide the flight and spraying operation of the UAV.

[0125] S55. During the path planning process, the fault spread area is taken as the priority target area for spraying to ensure timely coverage of the fault area.

[0126] Specifically, in some embodiments of this application, a fault propagation prediction model is used to predict a fault propagation area, and the boundary coordinates of the fault propagation area are determined based on the prediction results. When using the fault propagation prediction model, the model predicts the boundary coordinates of the fault propagation area to be (10, 20, 30) and (40, 50, 60). This indicates that the boundary coordinates of the fault propagation area are (x = 10 to 40, y = 20 to 50, z = 30 to 60). The obstacle distribution obtained through the 3D mapping model shows that there is an obstacle with a height of 50 within the range of (x = 10 to 40, y = 20 to 50, z = 30 to 60). Based on the above obstacle distribution results, a fast expanding random tree algorithm is used for path planning. The spraying area is divided into a series of discrete points at 1-meter intervals, resulting in a set of discrete points {(10, 20, 30), (11, 21, 31), ..., (40, 50, 60)}. The drone's starting point coordinates are (10, 20, 30), and the RRT algorithm is used for path planning. In each iteration, the RRT algorithm randomly samples a point and gradually expands the tree structure through rapid expansion. During expansion, obstacle distribution is acquired to ensure the generated path does not intersect with obstacles. Simultaneously, the shortest distance is prioritized to guide the RRT algorithm's expansion direction, generating the optimal spraying path. During expansion, the RRT algorithm selects points (20, 30, 40), (30, 40, 50), and (40, 50, 60) as key expansion points and successfully generates an optimal path from the starting point to the target point. The drone's flight and spraying operations are guided by the discrete points (20, 30, 40), (30, 40, 50), and (40, 50, 60) on the optimal path. During path planning, the fault diffusion area is prioritized for spraying to ensure timely coverage of faulty areas.

[0127] Furthermore, step S5 of this application also includes the following steps:

[0128] S56. Based on the turning radius of the UAV, optimize and smooth the path to obtain the initial path. Determine the turning radius limit of the UAV based on its size, maneuverability, and safety requirements.

[0129] S57. Use spline path interpolation to process the initial path, adjust the control points or parameters of the path, and during the smoothing process, adjust the curvature and turning angle of the path according to the turning radius limit of the UAV to ensure that the UAV turns smoothly without exceeding the limit.

[0130] S58. Establish a convolutional neural network model to analyze the spraying effect. Use known good spraying samples and uncovered area samples as inputs to the convolutional neural network model for supervised learning. The convolutional neural network model outputs the uncovered area.

[0131] S59. After the spraying is completed, input the image after spraying into the trained convolutional neural network model, output the uncovered area, that is, the part that is not fully covered due to insufficient spraying angle, determine whether there is an area that is not covered due to insufficient spraying angle, and if there is an area that is not covered due to insufficient spraying angle, optimize the spraying path.

[0132] Specifically, in some embodiments of this application, the UAV is 2 meters × 2 meters in size, with a maximum allowable turning radius of 5 meters, and safety requirements stipulate that the UAV's turning angle does not exceed 90 degrees during the turning process. The initial path is a straight segment that needs to be adjusted according to the UAV's turning radius limit. The initial path is a straight segment from point A to point B, with a path length of 10 meters. According to the turning radius limit, the UAV needs to maintain a turning radius of 5 meters during the turning process. Therefore, a turning point C needs to be inserted between point A and point B to allow the UAV to turn smoothly. According to the spline path interpolation algorithm, the path is adjusted by adjusting the control points or parameters of the path. By adjusting the path parameters, the path forms an arc with a radius of 5 meters at the turning point C. During the smoothing process, according to the UAV's turning radius limit, the curvature and turning angle of the path need to be adjusted. By adjusting the curvature of the path, the curvature of the UAV during the turning process is kept on an arc with a radius of 5 meters, and the turning angle does not exceed 90 degrees. After spraying, a convolutional neural network model was used to analyze the spraying coverage effect. The convolutional neural network model output the uncovered areas, indicating parts where the spraying angle was insufficient, resulting in incomplete coverage. Based on the results, the spraying path was optimized, and the control points or parameters of the path were adjusted to ensure that the spraying path performed touch-up spraying on the uncovered areas.

[0133] For step S6 of this application, it specifically includes the following steps:

[0134] S61. Use the fault propagation prediction model to predict the fault propagation area and determine the boundary coordinates of the fault propagation area.

[0135] S62. Determine the limit on the single flight time of the drone based on its battery capacity, flight speed, and endurance.

[0136] S63. Determine the processing priority of the fault area within the time constraint based on the area of ​​the fault area and the distance from the drone's starting point. The larger the spread area and the closer the distance to the obstacle, the higher the priority.

[0137] S64. Sort the faulty areas according to priority and determine the order of processing;

[0138] S65. Based on the sorted priorities, formulate a processing plan within the time constraint. Areas with higher priorities than the preset priority are arranged in the first flight painting plan, and areas with lower priorities than the preset priority are placed in the second flight plan. Based on real-time risk and energy consumption data, dynamically optimize the safety and efficiency of UAV painting operations.

[0139] Specifically, a fault propagation prediction model was used to predict a fault propagation region. The boundary coordinates of the fault propagation region are: top left corner (10, 20) and bottom right corner (30, 40). The single flight time limit for the UAV was determined. The UAV's battery capacity is 31250mAh, its flight speed is 10 m / s, and its endurance is 45 minutes. The single flight time limit for the UAV was calculated to be 45 minutes. Based on the area of ​​the fault region and its distance from the UAV's starting point, the processing priority of the fault regions within the time constraint was determined. Fault region A has an area of ​​100 square meters and is 20 meters from the UAV's starting point, while fault region B has an area of ​​80 square meters and is 30 meters from the UAV's starting point. Fault region A was determined to have a higher processing priority. The fault regions were sorted according to priority. There are two fault regions: region A has a priority of 2, and region B has a priority of 1. Based on priority, region A was placed at the front of the queue, and region B was placed at the back of the queue. Based on the sorted priorities, a processing plan within the time constraint was formulated. The time constraint is 20 minutes. Areas with a priority higher than the preset priority will be scheduled for the first flight painting plan, while areas with a priority lower than the preset priority will be placed in the second flight plan. Area A will be treated in the first flight, and area B will be treated in the second flight.

[0140] Furthermore, step S6 of this application also includes the following steps:

[0141] S66. Based on real-time risk and energy consumption data, the safety and efficiency of UAV spraying operations are dynamically optimized, and a comprehensive risk assessment model is constructed. The risk assessment formula is as follows:

[0142] R = a*(1 / E) + b*W + c*(1 / T) + d*D

[0143] Where R is the risk value, coefficients a, b, c, and d represent the weight of factors, 1 / E represents the reciprocal of the remaining battery energy (the lower the battery level, the higher the risk), W directly represents the wind speed (the higher the wind speed, the higher the risk), 1 / T is the reciprocal of the remaining working time (the shorter the time, the higher the risk), and D represents the distance to the nearest charging station (the farther the distance, the higher the risk).

[0144] S67. Construct an energy consumption prediction model. The energy consumption prediction formula is as follows:

[0145] C = a * P + b * S^2 + y * |V|

[0146] Where C is the energy consumption, a, b, and y are coefficients adjusted according to the equipment characteristics, P represents the spraying power consumption, S^2 represents the square of the flight speed, and |V| is the absolute value of the vertical speed change.

[0147] S68. Combining the outputs of the comprehensive risk assessment model and the energy consumption prediction model, the flight altitude, speed and spraying parameters of the UAV are adjusted in real time. When the risk assessment value exceeds the preset threshold or the energy consumption is higher than the preset value, the flight plan is automatically adjusted.

[0148] S69. By using the drone's sensors, wind speed, battery level, location, and flight time data are acquired in real time. The acquired data is fed back to the comprehensive risk assessment model and energy consumption prediction model in real time. When the risk is higher than the preset risk or the energy is lower than the preset energy, the emergency plan is activated.

[0149] Specifically, a comprehensive risk assessment model is constructed, with the risk assessment formula R = a*(1 / E) + b*W + c*(1 / T) + d*D, where coefficients a, b, c, and d represent the weights of different factors; E represents the remaining battery energy (lower battery level, higher risk); W represents wind speed (higher wind speed, higher risk); T represents the remaining operating time (shorter time, higher risk); and D represents the distance to the nearest charging station (greater distance, higher risk). A certain drone has a remaining battery energy E of 20%, a wind speed W of 25 m / s, a remaining operating time T of 10 minutes, and a distance D of 2 kilometers to the nearest charging station. The coefficients are set to a = 5, b = 3, c = 2, and d = 1. According to the risk assessment formula, R = 5*(1 / 0.2) + 3*25 + 2*(1 / 10) + 1*2 = 25 + 75 + 0.2 + 2 = 102.2. Therefore, the risk assessment value of this drone is 102.2. An energy consumption prediction model is constructed, with the formula C = a*P + b*S^2 + y*|V|, where a, b, and y are coefficients adjusted according to equipment characteristics, P represents spraying power consumption, S represents flight speed, and |V| represents the absolute value of vertical speed change. A certain UAV has a spraying power consumption P of 10W, a flight speed S of 15m / s, and an absolute value of vertical speed change V of 5m / s. Coefficients a, b, and y are set to 7, 2, and 1. Based on the energy consumption prediction formula, C = 7*10 + 2*(15^2) + 1*5 = 70 + 450 + 5 = 525. Therefore, the predicted energy consumption value for this UAV is 525. Combining the outputs of the comprehensive risk assessment model and the energy consumption prediction model, the UAV's flight altitude, speed, and spraying parameters are adjusted in real time. When the risk assessment value exceeds a preset threshold or the energy consumption is higher than a preset value, the flight plan can be automatically adjusted. The preset risk assessment threshold is 80, and the preset energy consumption is 400. If the risk assessment value is 102.2, exceeding the preset threshold of 80, the flight altitude, speed, and spraying parameters are adjusted to reduce the risk. If the predicted energy consumption value is 400, higher than the preset energy consumption of 525, the flight altitude, speed, and spraying parameters are adjusted to reduce energy consumption. The drone's sensors acquire real-time data on wind speed, battery level, location, and flight time, feeding this data back to the comprehensive risk assessment model and energy consumption prediction model for real-time calculation and evaluation. When the risk exceeds the preset risk level or the energy consumption falls below the preset energy level, emergency plans are activated, including automatic return to base, reducing flight altitude, or switching to a safe mode. Multiple emergency plans are designed to address different risk and energy conditions.

[0150] For step S7 of this application, it further includes the following steps:

[0151] S71. Based on the determined fault area priority, load the corresponding painting path into the UAV's flight control system;

[0152] S72. Select the appropriate nozzle type for the path line according to the priority area type. If the priority area is a plane, select a plane atomizing nozzle; if it is a curved surface, select a cone nozzle with deflection.

[0153] S73. Determine the nozzle flow rate based on the priority of the fault area;

[0154] S74. Select the optimal nozzle based on nozzle flow rate and nozzle type;

[0155] S75. Calculate the required spraying speed based on the size of the fault area and the nozzle flow rate;

[0156] S76. Determine the required spraying pressure based on the nozzle flow rate and the spraying speed.

[0157] In some specific embodiments of this application, a large transformer is being repaired by spray painting, and the fault area is determined to be a planar area. Based on the priority of the fault area, its corresponding spraying path needs to be loaded into the UAV's flight control system. Since the priority area type is planar, a planar atomizing nozzle with a suitable path shape is selected. Based on the size and priority of the fault area, the nozzle flow rate is determined to be 100 liters / hour. Based on the nozzle flow rate and nozzle type, the optimal nozzle is selected as a planar atomizing nozzle with a flow rate of 100 liters / hour. Based on the size of the fault area and the nozzle flow rate, the required spraying speed is calculated. If the area of ​​the fault area is 100 square meters and the nozzle flow rate is 100 liters / hour, then the required spraying speed is 1 square meter / hour. Based on the nozzle flow rate and spraying speed, the required spraying pressure is determined. If the nozzle flow rate is 100 liters / hour and the spraying speed is 1 square meter / hour, then the required spraying pressure is 100 kPa.

[0158] Step S8 of this application specifically includes the following steps:

[0159] S81. Based on the spraying pressure and speed parameters and the constraints of nozzle unobstructedness, determine the viscosity range of the coating, perform rotational viscosity testing on the coating, and obtain the viscosity data of the coating.

[0160] S82. Determine whether the paint passes smoothly through the nozzle for spraying. If the nozzle is blocked or not clear, it means that the paint viscosity is too high. Dilute or adjust the paint composition to reduce the paint viscosity. If there is overspraying and paint runoff, it means that the paint viscosity is too low. Thicken or adjust the paint composition to increase the paint viscosity.

[0161] S83. Based on the conductivity requirements during the spraying process, adjust the composition of the coating to make it conductive, and determine whether the adhesion of the coating meets the requirements. If it does not meet the requirements, adjust the composition of the coating to improve the adhesion of the coating.

[0162] S84. Based on the viscosity, conductivity, and adhesion of the coating, determine the final coating formula and select a drone spraying method that matches the viscosity of the coating.

[0163] For example, the viscosity range of the coating is determined based on spraying pressure and speed parameters, as well as nozzle unobstructedness constraints. With a spraying pressure of 3 bar, a spraying speed of 2 m / s, and nozzle unobstructedness requirements, the viscosity must be between 10-1000 cP. A rotational viscosity test is performed on the coating to obtain viscosity data, showing a viscosity of 300 cP. Based on the spraying pressure and speed parameters and nozzle unobstructedness constraints, it is determined whether the coating viscosity is within the required range. According to the parameters and test results, 300 cP is within the 10-1000 cP range, indicating that the coating can be sprayed through the nozzle. If the nozzle is clogged or obstructed, the coating viscosity is too high, requiring the addition of a thinner, such as a solvent, to reduce it. If overspraying and coating runoff occur, the coating viscosity is too low, requiring the addition of a thickener to increase it. Based on the conductivity requirements during the spraying process, the coating composition is adjusted to make it conductive, and conductive fillers, such as metal particles, are added to increase the coating's conductivity. Determine if the coating's adhesion meets the requirements. If not, adjust the coating composition to improve adhesion by adding adhesives or changing the resin formulation. Determine the final coating formulation based on its viscosity, conductivity, and adhesion. Based on the test results and requirements, select a drone spraying system with a matching coating viscosity. If the coating viscosity is found to be 300 cP during testing, select a spraying equipment suitable for that viscosity.

[0164] Furthermore, step S8 of this application also includes the following steps:

[0165] S85. The formula for adjusting the paint viscosity during UAV spraying based on real-time environmental data and predicting paint behavior is as follows:

[0166] P = a*T + b*H + c*V + d*L;

[0167] Where P represents the predicted spraying parameters, T represents the ambient temperature, H represents the ambient humidity, V represents the wind speed, L represents the viscosity of the coating, and a, b, c, and d are adjustment coefficients.

[0168] S86. Install environmental monitoring equipment on the drone to collect temperature, humidity and wind speed data in real time, and obtain real-time rheological property data of the coating, including viscosity and flowability.

[0169] S87. Using the paint behavior formula, adjust the spraying speed, pressure and nozzle angle based on real-time monitoring data. When the temperature rises or the wind speed increases, adjust the spraying pressure and speed to maintain the uniform distribution of the paint.

[0170] S88. Based on actual spraying test data, calibrate and optimize the parameters in the formula, and update the model regularly to adapt to new coating properties and environments.

[0171] For example, the paint behavior formula is P=5*T+3*H+2*V+1*L. Given an ambient temperature of 25 degrees Celsius, humidity of 65%, wind speed of 10 m / s, and paint viscosity of 5000 cps, the predicted spraying parameters are calculated as P=5*25+3*65+2*10+1*5000=125+165+20+500=5340. Based on these predicted parameters, the spraying speed, pressure, and nozzle angle are adjusted to maintain uniform paint distribution. If the temperature rises to 30 degrees Celsius, the spraying parameters can be recalculated as P=5*30+3*65+2*10+1*5000=1500+165+20+5000=5365. Based on the new parameters, the spraying speed, pressure, and nozzle angle are adjusted to adapt to the temperature change. By calibrating and optimizing the parameters in the formula using actual spraying test data, the optimal values ​​of adjustment coefficients a, b, c, and d are determined to adapt to different coating characteristics and environmental conditions. Regularly updating the model can maintain its accuracy and improve the coating spraying effect.

[0172] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A drone painting method based on path planning, characterized in that, include: S1. Obtain image data of overhead lines, segment the line images, and extract images of key parts of the overhead lines. S2. Match the extracted key part images with the standard key part images to determine the state of the key parts; S3. If there is a fault in a critical part, record the fault type and establish a three-dimensional mapping model, mark the fault location coordinates, and obtain the obstacle avoidance space distance around the fault. S4. Establish a fault propagation prediction model to predict the fault propagation area based on the fault location and fault type. S5. Based on the predicted fault spread range and the terrain of the spraying area, plan the UAV circuit spraying path and prioritize spraying the fault spread area. S6. Based on the limitations of the diffusion area and the single flight time of the UAV, determine the processing priority of the fault area within the time constraint, including: S65. Based on the sorted priorities, formulate a processing plan within the time constraint. Areas with higher priorities than the preset priorities are arranged in the first flight painting plan, and areas with lower priorities than the preset priorities are placed in the second flight plan. Based on real-time risk and energy consumption data, dynamically optimize the safety and efficiency of UAV painting operations. S66. Based on real-time risk and energy consumption data, the safety and efficiency of UAV spraying operations are dynamically optimized, and a comprehensive risk assessment model is constructed. The risk assessment formula is as follows: R = a*(1 / E) + b*W + c*(1 / T) + d*D Where R is the risk value, coefficients a, b, c, and d represent the weight of factors, 1 / E represents the reciprocal of the remaining battery energy (the lower the battery level, the higher the risk), W directly represents the wind speed (the higher the wind speed, the higher the risk), 1 / T is the reciprocal of the remaining working time (the shorter the time, the higher the risk), and D represents the distance to the nearest charging station (the farther the distance, the higher the risk). S67. Construct an energy consumption prediction model. The energy consumption prediction formula is as follows: C = a * P + b * S^2 + y * |V| Where C is the energy consumption, a, b, and y are coefficients adjusted according to the equipment characteristics, P represents the spraying power consumption, S^2 represents the square of the flight speed, and |V| is the absolute value of the vertical speed change. S68. Combining the outputs of the comprehensive risk assessment model and the energy consumption prediction model, the flight altitude, speed and spraying parameters of the UAV are adjusted in real time. When the risk assessment value exceeds the preset threshold or the energy consumption is higher than the preset value, the flight plan is automatically adjusted. S69. By using the drone's sensors, wind speed, battery level, location, and flight time data are acquired in real time. The acquired data is fed back to the comprehensive risk assessment model and energy consumption prediction model in real time. When the risk is higher than the preset risk or the energy is lower than the preset energy, the emergency plan is activated. S7. Determine the spraying path based on the fault handling priority, select the best nozzle, and set the spraying speed and spraying pressure. S8. When using a drone for live-line spraying, adjust the paint viscosity according to the spraying pressure and speed, check whether the nozzle is unobstructed, and after adjusting the paint, use a matching drone for live-line spraying.

2. The UAV painting method based on path planning according to claim 1, characterized in that, Step S1 includes: S11. Obtain an image dataset containing overhead lines by taking on-site photos with a drone. This image dataset contains images of overhead lines under different preset scenes and different preset lighting conditions. S12. Denoise, resize, and enhance the contrast of the acquired image; S13. The image is annotated to identify key parts, including the locations of towers, insulators and jumpers, and the annotated dataset is divided into a training set and a validation set. S14. Train the model using the training set and optimize the model parameters using cross-entropy loss; S15. Segment the overhead line images in the test set, label the segmented areas, and extract the segmented areas belonging to towers, insulators, and jumpers based on the label information.

3. The UAV painting method based on path planning according to claim 2, characterized in that, Step S2 includes: S21. Set a threshold to distinguish the state of key parts and obtain standard key part images, including sample images in normal state and sample images in fault state, and use the extracted key part images as the key part images to be detected. S22. Through image preprocessing, the standard key part image and the key part image to be detected are normalized and adjusted to the same size and contrast. S23. Input the preprocessed standard key part images into the convolutional neural network, output the part state category, and train the network model through backpropagation. S24. Input the image of the key part to be detected into the trained convolutional neural network model, set a threshold based on the matching similarity to determine the state of the key part, and output the state classification result of the key part to be detected. If the state of the key part is fault, output a fault prompt, the fault prompt including the fault type, the name of the key part and the coordinates of the key part.

4. The UAV painting method based on path planning according to claim 3, characterized in that, Step S3 includes: S31. Using the camera and lidar of the UAV, acquire images of the surrounding area of ​​the route, establish a three-dimensional mapping model of the terrain and building obstacles around the route, and acquire the GPS position data of the UAV and the relative positional relationship between the UAV and the target on the route. S32. Based on the location of the UAV, obtain the coordinate information of the faulty part, use the three-dimensional mapping model to calculate the coordinates of the faulty part on the line, and convert the coordinate information of the faulty part into the absolute coordinates of the line. S33. Mark the fault coordinates in the three-dimensional mapping model, and obtain the three-dimensional data of the surrounding environment of the line based on the absolute coordinates of the fault on the line.

5. The UAV painting method based on path planning according to claim 4, characterized in that, Step S4 includes: S41. Obtain historical fault data, including the location coordinates, type, environmental conditions, and spread range of the fault; S42. Preprocess the data, filter out abnormal data, and perform normalization. S43. Extract features that affect fault propagation based on the historical fault data. The features that affect fault propagation include fault type, current parameters and climate conditions. Construct a feature vector for the fault sample. S44. Establish a fault propagation prediction model, input the obtained fault characteristics, predict the fault propagation area, and output the predicted fault propagation area. S45. If the predicted fault propagation area does not match the actual situation, obtain the sample data after the propagation and iteratively optimize the model. S46. Mark the predicted fault propagation area on the three-dimensional mapping model.

6. The UAV painting method based on path planning according to claim 1, characterized in that, Step S7 includes: S71. Based on the determined fault area priority, load the corresponding painting path into the UAV's flight control system; S72. Select the appropriate nozzle type for the path line according to the priority area type. If the priority area is a plane, select a plane atomizing nozzle; if it is a curved surface, select a cone nozzle with deflection. S73. Determine the nozzle flow rate based on the priority of the fault area; S74. Select the optimal nozzle based on nozzle flow rate and nozzle type; S75. Calculate the required spraying speed based on the size of the fault area and the nozzle flow rate; S76. Determine the required spraying pressure based on the nozzle flow rate and the spraying speed.

7. The UAV painting method based on path planning according to claim 6, characterized in that, Step S8 includes: S81. Based on the spraying pressure and speed parameters and the constraints of nozzle unobstructedness, determine the viscosity range of the coating, perform rotational viscosity testing on the coating, and obtain the viscosity data of the coating. S82. Determine whether the paint passes smoothly through the nozzle for spraying. If the nozzle is blocked or not clear, it means that the paint viscosity is too high. Dilute or adjust the paint composition to reduce the paint viscosity. If there is overspraying and paint runoff, it means that the paint viscosity is too low. Thicken or adjust the paint composition to increase the paint viscosity. S83. Based on the conductivity requirements during the spraying process, adjust the composition of the coating to make it conductive, and determine whether the adhesion of the coating meets the requirements. If it does not meet the requirements, adjust the composition of the coating to improve the adhesion of the coating. S84. Based on the viscosity, conductivity and adhesion of the coating, determine the final coating formula and select a drone spraying method with matching coating viscosity. S85. The formula for adjusting the paint viscosity during UAV spraying based on real-time environmental data and predicting paint behavior is as follows: P = a*T + b*H + c*V + d*L; Where P represents the predicted spraying parameters, T represents the ambient temperature, H represents the ambient humidity, V represents the wind speed, L represents the viscosity of the coating, and a, b, c, and d are adjustment coefficients. S86. Install environmental monitoring equipment on the drone to collect temperature, humidity and wind speed data in real time, and obtain real-time rheological property data of the coating, including viscosity and flowability. S87. Using the paint behavior formula, adjust the spraying speed, pressure and nozzle angle based on real-time monitoring data. When the temperature rises or the wind speed increases, adjust the spraying pressure and speed to maintain the uniform distribution of the paint. S88. Based on actual spraying test data, calibrate and optimize the parameters in the formula, and update the model regularly to adapt to new coating properties and environments.

8. The UAV painting method based on path planning according to claim 1, characterized in that, Step S5 includes: S51. Use a fault propagation prediction model to predict the fault propagation area, and determine the boundary coordinates of the fault propagation area based on the prediction results. S52. Use a three-dimensional mapping model to analyze the distribution of obstacles in the spraying area; S53. Based on the obstacle distribution analysis of the fault propagation area and the spraying area, plan the UAV circuit spraying path; S54. Based on obstacle distribution factors, use a fast-expanding random tree to generate the optimal spraying path and use it as the key point of the spraying path to guide the flight and spraying operation of the UAV. S55. During the path planning process, the fault spread area is taken as the priority target area for spraying to ensure timely coverage of the fault area. S56. Based on the turning radius of the UAV, optimize and smooth the path to obtain the initial path. Determine the turning radius limit of the UAV based on its size, maneuverability, and safety requirements. S57. Use spline path interpolation to process the initial path, adjust the control points or parameters of the path, and during the smoothing process, adjust the curvature and turning angle of the path according to the turning radius limit of the UAV to ensure that the UAV turns smoothly without exceeding the limit. S58. Establish a convolutional neural network model to analyze the spraying effect. Use known good spraying samples and uncovered area samples as inputs to the convolutional neural network model for supervised learning. The convolutional neural network model outputs the uncovered area. S59. After the spraying is completed, input the image after spraying into the trained convolutional neural network model, output the uncovered area, that is, the part that is not fully covered due to insufficient spraying angle, determine whether there is an area that is not covered due to insufficient spraying angle, and if there is an area that is not covered due to insufficient spraying angle, optimize the spraying path.

9. The UAV painting method based on path planning according to claim 1, characterized in that, Step S6 includes: S61. Use the fault propagation prediction model to predict the fault propagation area and determine the boundary coordinates of the fault propagation area. S62. Determine the limit on the single flight time of the drone based on its battery capacity, flight speed, and endurance. S63. Determine the processing priority of the fault area within the time constraint based on the area of ​​the fault area and the distance from the drone's starting point. The larger the spread area and the closer the distance to the obstacle, the higher the priority. S64. Sort the faulty areas according to priority and determine the order of processing.

Images